AI Paper Atlas

Research map / Detection and segmentation / CNN features

Looking Locally: Object-Centric Vision Transformers as Foundation Models for Efficient Segmentation

Manuel Traub, Martin V. Butz

ICML 2026 · Detection and segmentation › CNN features · approach: Convolution-transformer backbones · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use fovea-like input patching with FLIP-Large instead of full-image SAM2-LFLIP · Small objects (<1% image area), averagemean IoU (%): 73.50 -> 81.19Improved
use FLIP-Large instead of SAM-HSAM-H · Average over 6 datasets (incl. ObjaScale)mean IoU (%): 75.41 -> 83.26Improved
use FLIP-Tiny (0.51M parameters) instead of SAM2-LSAM2-L · Average over 6 datasets (incl. ObjaScale)mean IoU (%): 75.87 -> 79.90Improved
use FLIP-Large instead of SAM2-LSAM2-L · Inference time (RTX4090)time (ms): 88.64 -> 44.13Improved
scale FLIP from Tiny to LargeFLIP · Average over 6 datasets (incl. ObjaScale)mean IoU (%): 79.90 -> 83.26Improved
scale FLIP from Tiny to LargeFLIP · KITTI-360mean IoU (%): 72.62 -> 70.72Hurt
scale FLIP_bb from Tiny to LargeFLIP_bb · Box prompt noise 20%mean IoU (%): 30.76 -> 26.34Hurt
use box-derived Gaussian prompts (FLIP_bb-Large) instead of SAM-H under 20% box noiseSAM-H · Box prompt noise 20%mean IoU (%): 47.43 -> 26.34Hurt
use FLIP_bb-Large with clean boxes instead of SAM-HSAM-H · Box prompt noise clean (subset)mean IoU (%): 75.36 -> 79.53Improved
use native Gaussian prompts (FLIP-Large) instead of box-derived (FLIP_bb-Large) (ablation)FLIP_bb-Large · Average over 6 datasets (incl. ObjaScale)mean IoU (%): 79.46 -> 83.26Improved
use FLIP-Tiny instead of SAM-H on ObjaScaleSAM-H · ObjaScalemean IoU (%): 73.76 -> 88.55Improved
use SAM2-L instead of SAM-H on ObjaScaleSAM-H · ObjaScalemean IoU (%): 73.76 -> 69.54Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy